1. Technical Field
The present invention relates to a gaze target determination device which determines the object at which a user is gazing from among a plurality of objects displayed on a screen.
2. Background Art
Generally, the position of gaze on a screen obtained from eye gaze direction of a user is compared with a position of an object displayed on the screen to determine the object at which the user is gazing (hereinafter, may be simply referred to as a “gaze target”). Examples of a method for detecting an eye gaze direction include a corneal reflex method that estimates the gaze direction utilizing the positions of a pupil and a light source image (Purkinje image) that is a reflection, on a cornea, of near-infrared light irradiated on an eye by a light emitting diode (LED). Another example of a method for detecting an eye gaze is a method that detects the gaze direction utilizing a three-dimensional eye model or the like based on a user image captured by a video camera, assuming that a three-dimensional vector connecting the center of an eyeball and the center of an iris is the gaze direction (see for example, Non Patent Literature 1).
However, in the corneal reflex method, infrared light needs to be irradiated on the eye area; and thus, the detection range of the eye gaze is limited to the range where the infrared light can be irradiated. Thus, the corneal reflex method has a limitation in the user's posture or standing position where the eye gaze direction can be detected.
In the method disclosed in Non Patent Literature 1, it is not possible to directly measure the center position of an eyeball, the radius of the eyeball, or the like; and thus, parameters such as the eye center position or the radius of the eye are modeled using facial features. The estimation of the gaze direction with use of the three-dimensional eye model has an advantage that the user can take any postures or standing positions freely. On the other hand, a disadvantage also exists where a relatively large error may occur between the line connecting the eyeball center and the iris center and the actual eye gaze due to an error in calibration for the user or an error in the facial feature detection. In order to reduce such errors, there are proposed methods that aim to improve the accuracy of the gaze detection using multiple cameras having different viewpoints. However, in this type of method, the estimation accuracy of the gaze direction is still not significantly improved.
As described, in the case where the user takes various postures freely to some extent, the estimated gaze direction includes a relatively large error. Thus, in order to accurately determine the gaze target, there are possible approaches which focus on the tracking capability of the eye gaze relative to a change in an object on a screen, instead of directly comparing the gaze position obtained from the gaze direction with the display position of an object (see Patent Literature 1 and Patent Literature 2).
The attention determination device disclosed in Patent Literature 1 moves display information Q displayed on a screen, when the gaze is directed on the screen. When the detected gaze movement direction corresponds with the movement of the display information Q, the attention determination device determines that the gaze tracked the movement of the display information Q.
The driver's condition determining device disclosed in Patent Literature 2 gives a driver a visual stimulus. The driver's condition determining device also determines whether or not the driver is in the conditions suited for driving, based on detection results of the driver's gaze direction, gaze duration time, and a gaze trajectory when the position to which the stimulus is given is moved.